[rollout,docs] fix: improve error message (#4682) and docstrings (#1345)

[yurekami]

Summary

This PR contains two contributions:

1. Fix for Issue #4682 - Informative error message for generate_sequences

• Problem: vLLMAsyncRollout.generate_sequences() raised a bare NotImplementedError, leaving users confused when running generation scripts
• Root cause: The vLLM SPMD (sync) mode was retired in PR [vllm] feat: retires vllm spmd mode in the codebase #4411, but the generation workflow (main_generation.py) still expects a synchronous generate_sequences() method
• Fix: Added an informative error message explaining:
  • Sync mode was retired in PR [vllm] feat: retires vllm spmd mode in the codebase #4411
  • Users should use the async server interface (vLLMReplica, AsyncLLMServerManager)
  • Alternative: use HFRollout for synchronous generation
  • Links to issue Bug of examples/generation/run_deepseek_v2_lite_math.sh #4682 for details
• Also updated generation.yaml config comments to document the limitation

2. Documentation improvement for Issue #1345 - Google-style docstrings in device.py

Standardized all function docstrings in verl/utils/device.py to follow Google-style documentation format:

• is_torch_npu_available(): Added detailed description and return type
• get_visible_devices_keyword(): Clarified purpose and return values
• get_device_name(): Improved description of supported devices
• get_torch_device(): Documented fallback behavior
• get_device_id(): Concise description with example
• get_nccl_backend(): Explained HCCL vs NCCL selection
• set_expandable_segments(): Added OOM context and Note section
• auto_set_ascend_device_name(): Documented NPU auto-configuration
• get_device_capability(): Added proper type hints and description

Fixes #4682
Contributes to #1345

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[gemini-code-assist]

Code Review

This pull request delivers two valuable improvements. Firstly, it significantly enhances the developer experience by providing a much more informative error message for the deprecated synchronous generate_sequences method in vLLMAsyncRollout. This will prevent confusion and guide users toward the correct asynchronous implementation. Secondly, the standardization of docstrings in verl/utils/device.py to the Google style guide is a great step towards better code clarity and maintainability. I have added one high-severity comment to address a regression in type hinting within device.py to ensure consistency with the improvements being made. Overall, this is a solid contribution.

[gemini-code-assist]

While the previous type hint -> any was incorrect (it should be -> Any), removing the type hint altogether is a regression, especially in a PR that aims to improve documentation and typing. To maintain consistency with the other functions in this file and improve code clarity for developers and static analysis tools, a correct return type hint should be provided.

The function returns a module object (e.g., torch.cuda), so types.ModuleType is the most appropriate type hint. You will need to add import types at the top of the file.

Suggested change

	def get_torch_device():
	def get_torch_device() -> "types.ModuleType":